K-means via scipy vs scikit-learn

Introduction

K-means helps group similar data points together. Using scipy or scikit-learn are two ways to do this in Python.

You want to find groups in customer data to offer personalized deals.

You want to organize photos by similar colors or features.

You want to simplify complex data by grouping similar items.

You want to compare how different tools perform the same task.

You want to learn how clustering works using popular Python libraries.

Syntax

SciPy

from scipy.cluster.vq import kmeans, vq

# data = your data array
centroids, distortion = kmeans(data, k)
cluster_labels, _ = vq(data, centroids)

scipy uses kmeans to find centers and vq to assign points.

scikit-learn uses KMeans class with fit and predict methods.

Examples

Using scipy to find 2 clusters and assign labels.

SciPy

from scipy.cluster.vq import kmeans, vq
import numpy as np

data = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])
centroids, distortion = kmeans(data, 2)
labels, _ = vq(data, centroids)
print('Centroids:', centroids)
print('Labels:', labels)

Using scikit-learn to do the same clustering with simpler code.

SciPy

from sklearn.cluster import KMeans
import numpy as np

data = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])
kmeans = KMeans(n_clusters=2, random_state=0).fit(data)
print('Centroids:', kmeans.cluster_centers_)
print('Labels:', kmeans.labels_)

Sample Program

This program shows how to run K-means clustering on the same data using both scipy and scikit-learn. It prints the cluster centers and labels for each method so you can compare.

SciPy

from scipy.cluster.vq import kmeans, vq
from sklearn.cluster import KMeans
import numpy as np

# Sample data: points in 2D space
data = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])

# Using scipy
centroids_scipy, distortion = kmeans(data, 2)
labels_scipy, _ = vq(data, centroids_scipy)

print('Scipy K-means results:')
print('Centroids:', centroids_scipy)
print('Labels:', labels_scipy)

# Using scikit-learn
kmeans_sklearn = KMeans(n_clusters=2, random_state=42).fit(data)
print('\nScikit-learn K-means results:')
print('Centroids:', kmeans_sklearn.cluster_centers_)
print('Labels:', kmeans_sklearn.labels_)

Important Notes

Scipy's kmeans returns centroids and distortion (how good the clusters are).

Scikit-learn's KMeans class is easier to use and has more options like initialization methods.

Both methods give similar results on simple data but scikit-learn is preferred for real projects.

Summary

K-means groups data points into clusters based on similarity.

Scipy requires two steps: find centroids, then assign labels.

Scikit-learn combines these steps and offers more features.